The goal of this longitudinal, multilevel study was to develop a better understanding of poor renter households' mobility patterns by identifying the relative importance of individual and contextual variables. Variability in neighborhood poverty rates (NPR) was analyzed for 1564 poor, renter households living in 179 metropolitan statistical areas (MSAs) across the continental U.S. during the 1990s. Household heads were typically black (73%), middle age (mean=37 years) females (59%) who had 12 or fewer years of education (77%). Each household completed three to nine Panel Study of Income Dynamics (PSID) surveys. Using geocodes, census data were linked with survey data to provide information about the NPR and metropolitan opportunity structure at each survey occasion.

Multilevel modeling was used to analyze this hierarchically-structured data (measurement occasions nested within households nested within MSAs). While 58% of variability in outcomes was due to between-household differences, 15% was due to between-MSA differences (the remainder was between-measurement occasion variability). Each of the three blocks of predictors significantly improved the model: individual decisions (work, housing, fertility and marriage), personal characteristics (race, age, gender and education) and MSA characteristics (segregation, housing, labor market and area poverty conditions).

Controlling for other predictors, race was the most important predictor, increasing a black household's NPR by over ten points and interacting with several other predictors. Being black amplified the negative effect of having more children, weakened positive effects of increased income and a better MSA opportunity structure, and interacted with MSA segregation to the disadvantage of black households. Increased education lowered the NPR. Across income levels, the average white household lived in a non-poor neighborhood while the average black household had an NPR nearly twice as high.

Living in public housing was associated with a 4.7 percentage point differential in NPR (compared to no assistance). Other forms of government-assisted housing also increased the NPR, but by less than one percentage point. Mobility lowered the NPR, as did becoming a homeowner.

Individual choices made a difference, but characteristics individuals were born with amplified or diminished effects of their efforts. The NPR was further influenced by housing type, tenure and mobility. Most importantly, metropolitan context mattered.